Allow linlog to be used.

[govertb]

[jeremiegarrigues]

I tried to implement a version of the linlog by modifying the calculation of attraction forces in RPFA2Kernels.cu, without success. Adding a non-linear function causes a memory fault on line 234 of the RPGPUForceAtlas2.cu file during the execution (cudaCatchError(cudaDeviceSynchronize())). Could you give me an indication on how to implement the linlog version of ForceAtlats2 for the CUDA implementation? Thank you in advance.

[jeremiegarrigues]

It was actually a cast problem. Forces, although declared in const float, only support integers. However, after applying logarithms to the repulsive forces, the result is not very convincing. The Noack model should not be reduced to a simple logarithm on a part of the applied forces. If you have any advice.

[govertb]

Hi, Thanks for sharing your findings and sorry for the late response. I'll try to look into this in the coming weeks... Casting forces to int shouldn't be necessary... I expect that something else is causing the memory fault (maybe related to #10).